Tuning the Parameters: A Maritime-Tuned Machine Learning Course

Vincenzo Antonio Ventricelli; Paul M. Kump; Van-Hai Bui

Download Paper | Permalink

Conference: 2023 ASEE Annual Conference & Exposition
Location: Baltimore , Maryland
Publication Date: June 25, 2023
Start Date: June 25, 2023
End Date: June 28, 2023
Conference Session: Ocean & Marine Engineering Division Technical Session 2
Tagged Division: Ocean and Marine Division (OMED)
Page Count: 19
DOI: 10.18260/1-2--44523
Permanent URL: https://peer.asee.org/44523
Download Count: 265

Paper Authors

biography

Vincenzo Antonio Ventricelli SUNY Maritime College

visit author page

Vincenzo Ventricelli is an undergraduate student and student researcher at the State University of New York Maritime College pursuing a bachelor's degree in Electrical Engineering and a USCG Unlimited License. The focus of his current research is the applications of machine learning in the maritime industry, including the use of maritime-related datasets in the classroom. In addition to machine learning, he has a deep interest in other electrical engineering-related topics such as communications theory, control engineering, and power distribution.

visit author page

biography

Paul M. Kump SUNY Maritime College

visit author page

Dr. Paul M. Kump joined SUNY Maritime College in 2012 and is Associate Professor of Electrical Engineering. His research interests are in the areas of machine learning (ML), signal processing, and alternative teaching strategies. Dr. Kump has developed intelligent software-defined radio for the US Navy in electronic warfare, nuclear material detection algorithms for the US government at US seaports, and crime prediction software for the Chicago Police Department. He recently collaborated with Mount Sinai Hospital to create smart software for automatic error detection in patient radiation therapy treatment plans. In his spare time, Dr. Kump works to combine his research with his love of electronic music performance, teaching machines the craft of songwriting. With extensive course and curriculum design experience, Dr. Kump is continuously committed to developing engineering programs that best prepare students for the ever-changing demands of industry leaders. His teaching interests include online and HyFlex education, as well as classroom flipping and education research-based tasks. He created Maritime College’s ENGR 396 Machine Learning course and has been recognized by Open SUNY for excellence in online teaching, pioneering the School of Engineering’s online course offerings.

visit author page

biography

Van-Hai Bui

visit author page

Dr. Van-Hai Bui received his B.S. degree in Electrical Engineering from the Hanoi University of Science and Technology in Vietnam in 2013 and his Ph.D. degree from Incheon National University in South Korea in 2020. From 2021 to 2022, he was a Research Investigator and Lecturer in the Department of Electrical and Computer Engineering at the University of Michigan-Dearborn. Currently, he is an Assistant Professor in the School of Engineering, Department of Electrical Engineering at the State University of New York (SUNY), Maritime College. His research interests include energy management systems, applications of reinforcement learning in power and energy systems, and microgrid operations.

visit author page

Download Paper | Permalink

Abstract

In machine learning (ML) education, the choice of which datasets to utilize for student assignments and projects is critical for student success and meeting course learning outcomes. Poorly chosen datasets leave students disinterested and questioning the applicability of ML in real-world situations specific to their intended endeavors post academia. Additionally, some datasets demand much effort for preprocessing and a steep learning curve for understanding, which detracts from the ML experience and leaves students frustrated. As maritime and marine engineering programs expand to include ML in their curricula with the plan of addressing industry trends in, for example, autonomy and defense, it is important to calibrate the ML course accordingly with relevant datasets and assignments. We develop a maritime-specific course in undergraduate ML (taken in the sixth semester) with the purpose of engaging students whose interests include maritime and marine industries. In support of our course, we compose several maritime-specific machine learning mini-projects employing the popular and convenient Google Colab platform and make them publically available through the GitHub repository. A hybrid of programming and report writing, each mini-project utilizes the same publically available maritime-related dataset—one that requires little preprocessing and, we show, is conducive for demonstrating many of the concepts vital to classical ML, as well as some topics in deep learning. Using the same dataset for many assignments fosters a feeling of student comfortability, promotes comparing the performances of different ML algorithms, and provides a low barrier of entry after the initial assignment. Our paper is both a detailed syllabus of a first course in maritime-focused ML and a how-to guide for effective use of the mini-projects we have developed. Going further, it is a solution to the mini-projects, as it reports on ML algorithms’ performances, how the choices of key tuning parameters affect said performances, and how and why algorithms perform the way they do. Concluding the paper is a student reflection authored by a US Coast Guard license student in engineering to offer instructors a unique student perspective and insight into the efficacy of the course design. Our hope is that colleagues interested in teaching a similar course at their own institutions can adopt our methods, and thereby reduce their preparation work and increase student engagement.

Citation
Format

Ventricelli, V. A., & Kump, P. M., & Bui, V. (2023, June), Tuning the Parameters: A Maritime-Tuned Machine Learning Course Paper presented at 2023 ASEE Annual Conference & Exposition, Baltimore , Maryland. 10.18260/1-2--44523

TY  - CPAPER
AB  - In machine learning (ML) education, the choice of which datasets to utilize for student assignments and projects is critical for student success and meeting course learning outcomes.  Poorly chosen datasets leave students disinterested and questioning the applicability of ML in real-world situations specific to their intended endeavors post academia.  Additionally, some datasets demand much effort for preprocessing and a steep learning curve for understanding, which detracts from the ML experience and leaves students frustrated.  As maritime and marine engineering programs expand to include ML in their curricula with the plan of addressing industry trends in, for example, autonomy and defense, it is important to calibrate the ML course accordingly with relevant datasets and assignments.   
We develop a maritime-specific course in undergraduate ML (taken in the sixth semester) with the purpose of engaging students whose interests include maritime and marine industries.  In support of our course, we compose several maritime-specific machine learning mini-projects employing the popular and convenient Google Colab platform and make them publically available through the GitHub repository.  A hybrid of programming and report writing, each mini-project utilizes the same publically available maritime-related dataset—one that requires little preprocessing and, we show, is conducive for demonstrating many of the concepts vital to classical ML, as well as some topics in deep learning.  Using the same dataset for many assignments fosters a feeling of student comfortability, promotes comparing the performances of different ML algorithms, and provides a low barrier of entry after the initial assignment. 
Our paper is both a detailed syllabus of a first course in maritime-focused ML and a how-to guide for effective use of the mini-projects we have developed.  Going further, it is a solution to the mini-projects, as it reports on ML algorithms’ performances, how the choices of key tuning parameters affect said performances, and how and why algorithms perform the way they do.  Concluding the paper is a student reflection authored by a US Coast Guard license student in engineering to offer instructors a unique student perspective and insight into the efficacy of the course design.  Our hope is that colleagues interested in teaching a similar course at their own institutions can adopt our methods, and thereby reduce their preparation work and increase student engagement.

AU  - Vincenzo Antonio Ventricelli
AU  - Paul M. Kump
AU  - Van-Hai Bui
CY  - Baltimore , Maryland
DA  - 2023/06/25
PB  - ASEE Conferences
TI  - Tuning the Parameters: A Maritime-Tuned Machine Learning Course
UR  - https://peer.asee.org/44523
DO  - 10.18260/1-2--44523
ER  -